The drive would not stay still
Windows 98’s Disk Defragmenter could begin again when the contents of the drive it was organizing changed. Microsoft documented that behavior directly in its Knowledge Base. The same note says the computer remained usable while defragmentation ran, but operated more slowly. Being allowed to keep working did not mean that doing so left the maintenance task undisturbed.
The conflict came from the job the tool was performing. Files could occupy multiple locations on a hard disk, a condition called fragmentation. Defragmenter reorganized file storage into contiguous blocks so the parts were closer together. A change to the drive’s contents altered the material being organized, and the old utility responded by restarting its process.
That explanation concerns changes to storage. The documentation does not say that moving the pointer was itself enough to trigger a restart. An apparently quiet desktop could still have software activity, while a visible action need not necessarily change the relevant drive. The important boundary was the drive’s contents, not whether the person watching had touched the mouse.
The separate problem at ten percent
Not every disappointing progress bar meant the same thing. Microsoft’s description of the Windows 98 tool explains that it normally performed program optimization during the first ten percent of the run. The time required depended on the programs being optimized and their sizes. A machine could look stuck while this work continued.
An older version of Knowledge Base article Q186978 describes that case specifically. It says the hard disk could remain active while the indicator sat at about ten percent. During program optimization the utility moved a smaller amount of clusters than usual, slowing the apparent progress. Microsoft classified that behavior as intentional.
Two symptoms, two explanations
A restart after a drive change and a slow initial optimization phase therefore belong to different parts of the utility’s behavior. One involved a moving target; the other involved work that the percentage display represented poorly. Treating both as a failed disk loses the distinction recorded in the contemporary support articles.
The old experience was a negotiation between ordinary computing and a maintenance program that benefited from a quiet drive. Watching the blocks rearrange did not guarantee a steady journey from the first percentage to the last.